EP3640946A1 - Multivariatischer ansatz für die auswahl der biologischen zelle - Google Patents
Multivariatischer ansatz für die auswahl der biologischen zelle Download PDFInfo
- Publication number
- EP3640946A1 EP3640946A1 EP18200386.3A EP18200386A EP3640946A1 EP 3640946 A1 EP3640946 A1 EP 3640946A1 EP 18200386 A EP18200386 A EP 18200386A EP 3640946 A1 EP3640946 A1 EP 3640946A1
- Authority
- EP
- European Patent Office
- Prior art keywords
- candidate cells
- cells
- sets
- data
- target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16B—BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
- G16B40/00—ICT specially adapted for biostatistics; ICT specially adapted for bioinformatics-related machine learning or data mining, e.g. knowledge discovery or pattern finding
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/048—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators using a predictor
Definitions
- the present application relates to cell, media, and process condition selection. More particularly, the application relates to/selecting at least one set of target cells from multiple sets of candidate cells.
- the cells may be clones.
- the candidate cells may be from a pool of heterogeneous groups of cells.
- the cells may be transfected or transformed. It may be desirable to scale up from a first scale (microscale) to a second scale (macroscale) one or more orders of magnitude greater than the first scale, and for the target cells to be stable throughout a manufacturing life cycle, e.g., a drug manufacturing life cycle.
- the target cells may need to meet targets corresponding to predetermined product quality attributes or predetermined key performance indicators.
- the cells may be used as part of (or to host) a chemical, pharmaceutical, biopharmaceutical and/or biological product. In some cases, it may be desirable to identify problematic heterogeneities in the candidate cells or in parent cells.
- Processes may generate huge amounts of data, particularly with respect to process parameters and/or conflicting (e.g., inversely correlated) product quality attributes. Problems regarding processing large amounts of data may be particularly acute in the case of a process control device capable of controlling, at least partly in parallel, processes in multiple vessels.
- the process control device may be capable of managing processes in 12 vessels, 24 vessels, or 48 vessels.
- Each vessel may be a bioreactor.
- the vessels may be microscale vessels.
- the vessels may have a volume of less than 5 L, less than 1 L, or less than 500 ml. More particularly, the vessels may have a volume of between 1 ml and 500 ml.
- Data analysis from experiments involving a process control device may involve multiple technicians spending many hours or days analyzing data in different data formats, possibly using a spreadsheet program such as Microsoft Excel. Results of the analysis may be inconsistent and/or subjective. In particular, it may be difficult to evaluate time series data of process parameters. Further, product quality attributes may be conflicting (e.g., inversely correlated), such that it is not possible to optimize two product quality attributes for the same process.
- Another problem is that data collected from a plurality of processes, particularly when the data is received at a process control device (e.g., some data is generated by the device itself and other data is generated by external analytic devices), is the presence of missing or unreliable data points. Such missing or unreliable data may lead to candidate cells being incorrectly evaluated and/or selection of an inefficient set of target cells.
- data collected from a plurality of processes may be stored in multiple or many data files.
- the data files may be spreadsheet files, e.g., Microsoft Excel files.
- macros are used to help sort and graph the data. It may be challenging to determine how to use all the data, particularly time series data. It may be challenging to determine which process outputs to consider, or what characteristics the process outputs should have. Further, when the originally selected characteristics cannot be met by any set of candidate cells, it may be challenging to determine a new set of characteristics in order to arrive at a selection of target cells. In addition, it may be a problem to prioritize process outputs without discarding one in favor of another. It may also be a problem to minimize the number of iterations of the process of selecting the set of target cells.
- a computer implemented method for selecting at least one set of target cells from multiple sets of candidate cells comprises receiving data collected from a plurality of processes, wherein each of the processes produces a distinct set of candidate cells.
- the received data includes values of process parameters and process outputs of the processes, each of the process outputs being a product quality attribute or a key performance indicator for selecting the target cells.
- the method further comprises correlating the received data, as well as receiving a selection of the process parameters and a selection of the process outputs.
- the method further comprises receiving multivariate evaluation criteria for the selected process parameters and/or the selected process outputs, the multivariate evaluation criteria including one or more of the following:
- Each prioritization range may specify an allowable span of values, e.g., 3 to 10.
- the prioritization range may also include an extremum, e.g., to indicate whether a higher or lower value within the range is preferred.
- Each prioritization target is an extremum (maximum or minimum), and/or a target value.
- a prioritization target may include an extremum (e.g., maximum) and a target value (e.g., 4.5).
- the prioritization target may include an extremum without a target value, or a target value without an extremum.
- Each of the processes produces a distinct set of candidate cells may be understood to mean that each one of the processes produces a different set of candidate cells. For example, if there are 18 processes, then 18 different sets of candidate cells are produced. Each set of candidate cells may be unique with respect to the other sets of candidate cells.
- process output may be considered an umbrella expression that covers both product quality attribute and key performance indicator.
- variable may be used to refer to either a process output or a process parameter.
- a set of cells may be a group or collection of cells having the same type (e.g., a group of CS1_1 clones may be a set of cells).
- a set of cells may be identified by a name (e.g., unique identifier) or location (e.g., culture station and/or vessel number) in a process control device.
- a name e.g., unique identifier
- location e.g., culture station and/or vessel number
- CS1 may be used to refer to cells located in a vessel in culture station 1 of the process control device.
- the digits following the "_” may identify individual vessels.
- CS1_1 may refer to the first vessel in culture station 1 and CS2_3 may refer to the third vessel in culture station 2.
- the method further comprises calculating, via a multivariate selection function, scores for each one of the sets of candidate cells from the correlated data according to the multivariate evaluation criteria.
- the method further comprises ranking the sets of candidate cells according to the scores, and selecting at least one of the sets of candidate cells as the target cells using the ranking.
- the target cells may be for use in (or to host) a chemical, pharmaceutical, biopharmaceutical and/or biological product. More particularly, the target cells may host the product, or the target cells may be the product.
- Each of the processes may be carried out by a process control device.
- the processes may be controlled by a single process control device, such that the processes are performed at least partly or entirely in parallel, or the processes may be carried out by additional process control devices.
- the processes may be carried out on a microscale. Alternatively, a portion of the processes may be carried out on a microscale and a portion of the processes may be carried out a macroscale. As another alternative, all the processes may be carried out on a macroscale.
- microscale may refer to vessels with a volume (i.e., working volume) as discussed above.
- Macroscale may refer to vessels having a volume (i.e., working volume) of greater than 1 L. More particularly, the microscale may be less than 250 ml and the macroscale may be greater than 3 L.
- the process parameters may include set points, flow rates, feed characteristics, initial conditions, online measurements, and offline measurements.
- set points include temperature, pH, dissolved oxygen, stirring speed.
- Flow rates may include air flow, carbon dioxide, oxygen, and acids/base.
- Nutrient characteristics may include an initial nutrient feed day, nutrient feed volume, and feed additions.
- Initial conditions may include seeding density and osmolarity.
- Online measurements may include temperature, pH, dissolved oxygen, volume of fluid in the vessel in which the process is being carried out.
- Offline measurements may include glucose, lactate, viable cell density (VCD), amino acid levels, monoclonal antibody concentration.
- a process output may be determined at the end of a corresponding process.
- a process parameter may be cell viability (e.g., measured during the process) and a process output may be the final cell viability (at the end of the process).
- Process outputs may include one or more of the following: a total quantity of cells, quantity of cells of per unit volume of input fluid, a chemical composition of the cells, a purity, amount of cell debris, amount of shear damage or chemical damage, starting material cost, energy cost for the process, product concentration, specific productivity, a profile describing the corresponding set of candidate cells (e.g., a glycan profile, a spectral profile), cell viability.
- selected process outputs and corresponding prioritization targets may be specified as follows:
- the process output is the profile of the set of candidate cells and the distance is an evaluation criterion received for the process output.
- the cells may be at least one of the following: a cell line, a cell strain, a clone.
- the multiple sets of candidate cells may form a heterogeneous pool of cells. More particularly, the multiple sets of candidate cells may form a heterogeneous transfection pool.
- the values of the process parameters may include time series values.
- the process parameters may have been controlled (controlled process parameters) and/or measured (measured process parameters) during each of the processes. Measurements may have been carried out online, at line, or offline.
- offline, atline and online may refer to the frequency at which fluid in a vessel (e.g., bioreactor) is monitored, e.g., by performing monitoring steps such as sampling the fluid.
- the fluid may contain a set of candidate cells.
- offline may also indicate that analysis of monitoring results is, at least in part, performed in a laboratory. For example, a sample obtained via offline monitoring may be transferred to a laboratory for time delayed laboratory analysis. Offline measurements may be carried out less than once per hour, e.g., twice per day.
- Atline measurements may be performed at a frequency similar to offline measurements. Atline measurements may involve analyzing an extracted sample in closer proximity to the vessel in comparison to offline measurements.
- Online measurements may be carried out with greater frequency than atline or offline measurements. For example, online measurements may be performed more than once per hour, more than three times per hour, or about sixty times per hour. Online measurements may be carried out in-situ or ex-situ. In-situ measurements might not involve removing a sample from the vessel. Instead, a sensor (e.g., temperature or pH sensor spot) may be directly inserted into the vessel or separated from the vessel by a wall. Another possible in-situ configuration involves a sampling loop with one online sensor, or a non-destructive online analyzer and return of a sample to the vessel after analysis. In online ex-situ measurements, the sample may be transported to an online analyzer and does not return to the vessel after analysis.
- a sensor e.g., temperature or pH sensor spot
- the described approach may be used to select media for cell cultivation or to set conditions of the process.
- the received data may include substantially all data from each of the processes.
- the received data may include values for each controlled process parameter and values for each measured process parameter. For example, if temperature within a vessel is measured throughout the process, then the received data may include all temperature measurements collected during the course of the process.
- the method may further comprise identifying whether the received data for any one of the processes is incomplete, wherein one of the processes is identified as having incomplete data when data is not collected during a portion of the process.
- the method may further comprise predicting values for the incomplete data using at least one multivariate technique.
- the multivariate technique may include partial least squares regression or interpolation.
- mechanistic modeling may also be used.
- the correlating may include verifying and correcting values of the data.
- the correcting may comprise revising or excluding values that violate one or more known metabolic dependencies.
- the dependencies may be ratios.
- the method may further comprise applying mechanistic modeling to the received data to obtain additional values of the process parameters and/or additional process outputs.
- the method may further comprise supplementing the received data with the additional values of the process parameters and/or the additional process outputs.
- At least one mechanistic model may be used to fit process parameter values and process outputs (possibly in the form of a process trajectory or cell growth profile) to monod growth kinetics.
- the mechanistic model may be used for data smoothing and/or for filling in missing data, as well as other beneficial applications. For example, given an example kinetic growth model, the maximum growth rate, death rate and other influences on growth can be determined from selected process outputs forming a process trajectory or cell growth profile.
- the mechanistic model and estimated values from the mechanistic model can be used to smooth process parameter values and process outputs. Smoothing may involve identifying and excluding measurement noise (i.e., inaccurate measurements). For example, viable cell density measurements may have an error of +/- 10%.
- the smoothing may involve excluding the erroneous measurements, e.g., process parameter values or process outputs.
- this approach provides meaningful information in explicit identification of growth, inhibition and death rates and can be used to calculate process outputs, such as maximum viable cell density or metabolite concentration (e.g., protein titer).
- maximum viable cell density or metabolite concentration e.g., protein titer
- the method may further comprise excluding, from the correlated data, data received from ones of the processes according to exclusion criteria. If at least one of the selected process outputs has a corresponding acceptability range, then the exclusion criteria may include the corresponding acceptability range for the at least one of the selected process outputs.
- the evaluation criteria may further comprise:
- the profile describing the process outputs may be a glycan profile or a spectral profile.
- the spectral profile may be a spectral line.
- the evaluation criteria may include a specified profile (e.g., a specified glycan profile and/or a specified spectral profile).
- the specified profile may correspond to (i.e., describe) a set of reference cells and may be used for comparison with the sets of candidate cells.
- the evaluation criteria may be specific to a set of cells or groups of different cells.
- the method may further comprise displaying the correlated data for the selected process parameters and/or the selected process outputs, comprising, displaying correlation patterns for the glycan profiles of the sets of candidate cells. Displaying the correlation pattern for the glycan profiles may involve displaying some or all final glycan measurements for one set of candidate cells and/or combining (e.g., via principal component analysis) glycan profiles for multiple sets of candidate cells.
- the selection function may include an objective function, particularly a cost function.
- the objective function may be used to calculate a distance (e.g., a Euclidian distance) between different sets of cells. The distance may be calculated based on orthogonal components derived from the selected process parameters and selected process variables.
- the objective function may provide a distance between one of the sets of candidate cells and a set of target values, e.g., from a set of reference cells.
- the output of the objective function may reflect the difference between the one of the sets of candidate cells and the set of target values.
- the output of the objective function may reflect a combination of distances according to the selected process parameter values and selected process outputs for the one of the sets of candidate cells and the set of target values (e.g., reference cells).
- the selection function may include (possibly in addition to the objective function) at least one magnifying function (also referred to as a penalty function).
- the magnifying function may magnify a distance between values (e.g., between values associated with one of the sets of candidate cells and target values).
- Each of the prioritization ranges and/or targets may have an associated magnifying function.
- the magnifying function may be non-linear a nonlinear polynomial function, e.g., exponential or quadratic. In some cases, a logarithmic function (e.g., a function of the natural logarithm or Euler's number) may be used.
- the magnifying functions may be the same for all prioritization targets or ranges. This may have the advantage of making it easier to prioritize targets or ranges by weight.
- magnifying functions may differ depending on the prioritization target or range. In particular, different magnifying functions may be used depending on the importance of the corresponding target.
- an initial distance may be calculated using the objective function and then magnified using the magnifying function. After magnification, the distance may be modified via a weight. In particular, the distance may be multiplied by the weight.
- the magnifying function and the weights provides additional flexibility, however, either one or both may be omitted for particular prioritization ranges/targets or entirely omitted from the evaluation. In particular, it may be possible to provide optimal or at least adequate cell selection without use of the magnifying function and/or the weights.
- Outputs of the selection function for each of the selected process parameters and/or the selected process outputs may be combined (e.g., summed) to calculate the score for the set of candidate cells.
- a computer program comprising computer readable instructions.
- the instructions when loaded and executed are on a computer system, cause the computer system to perform operations according to the method described above.
- the computer program may be implemented in the form of a computer program product, possibly (tangibly) embodied in a computer readable medium.
- a process control device for selecting at least one set of targets cells from multiple sets of candidate cells.
- the device comprises a plurality of vessels, each of the vessels being configured to contain fluid including one of the sets of candidate cells.
- the device further comprises a robot capable of addressing each of the vessels, dispensing fluid to each of the vessels, and extracting samples of fluid form each of the vessels.
- the device also comprises a controller operable to control, at least partly in parallel, conditions in each of the vessels.
- the controller is further operable to receive data collected from a plurality of processes, wherein each of the processes produces a distinct set of candidate cells.
- the received data includes values of process parameters and process outputs of the processes, each of the process outputs being a product quality attribute or a key performance indicator for selecting the target cells.
- the controller is further operable to correlate the received data, as well as to receive a selection of the process parameters, and a selection of the process outputs.
- the controller is further operable to receive multivariate evaluation criteria for the selected process parameters and/or the selected process outputs.
- the multivariate evaluation criteria include one or more of the following:
- the controller is further operable to calculate, via a multivariate selection function, scores for each one of the sets of candidate cells from the correlated data according to the multivariate evaluation criteria.
- the controller is further operable to rank the sets of candidate cells according to the scores, and select at least one of the sets of candidate cells as the target cells using the ranking.
- only one of the sets of candidate cells may be selected as the target cells.
- multiple sets of candidate cells e.g., 2-5) may be selected as target cells using the ranking.
- the subject matter described in this application excludes treatment of the human or animal body by surgery or therapy, and diagnostic methods practiced on the human or animal body.
- the subject matter described in this application can be implemented as a method or on a device, possibly in the form of one or more computer programs (e.g., computer program products). Such computer programs may cause a data processing apparatus to perform one or more operations described in the application.
- computer programs e.g., computer program products
- the subject matter described in the application can be implemented in a data signal or on a machine readable medium, where the medium is (tangibly) embodied in one or more information carriers, such as a CD-ROM, a DVD-ROM, a semiconductor memory, or a hard disk.
- the subject matter described in the application can be implemented as a system including a processor, and a memory coupled to the processor.
- the memory may encode one or more programs to cause the processor to perform one or more of the methods described in the application.
- Further subject matter described in the application can be implemented using various machines.
- Fig. 1 shows steps that may be carried out in a method for selecting at least one set of target cells from multiple sets of candidate cells.
- step S101 data collected from a plurality of processes is received. Each of the processes produces a distinct set of candidate cells.
- the received data includes values of process parameters and process outputs of the processes.
- Each of the process outputs is a product quality attribute or a key performance indicator for selecting the target cells.
- the sets of candidate cells are clones and the target cells are the best clone in the sets of candidate clones. It should be understood that although the following example is described in the context of clones, it is applicable to other types of cells.
- the plurality of processes are carried out in parallel via a process control device (shown in Figs. 8 and 9 and discussed in more detail below). More specifically, there are 24 processes carried out in 24 vessels. Each vessel has about 11 to 15 ml working volume. Data from offline measurements (e.g., glycan measurements, glucose, lactate, viable cell density (VCD), amino acid levels, monoclonal antibody concentration) is received at the process control device.
- offline measurements e.g., glycan measurements, glucose, lactate, viable cell density (VCD), amino acid levels, monoclonal antibody concentration
- the received data is organized into two separate tables (not shown) for 24 candidate clones.
- a first table is a process data table for seven process parameters including cell viability, process concentration, and specific productivity. Process parameter values were obtained over the course of several days taking one measurement per day.
- the second data table is a quality data table.
- the quality data table includes process outputs, specifically, one glycan profile having 13 separate process outputs (i.e., measurements) for each clone and a calculated distance from a target profile.
- Step S101 may also include correlating the received data.
- Step S101 may also include receiving a selection of the process parameters and a selection of the process outputs.
- the selection may be represented by the data stored in the process data table (i.e., the selection of process parameters) and the quality data table (i.e., the selection of process outputs).
- multivariate evaluation criteria for the selected process parameters and/or the selected process outputs is received.
- the multivariate criteria includes the following four prioritization targets:
- the distance specified for the quality prioritization target may be a Euclidian distance, calculated according to the formula specified in the context of S105.
- the prioritization targets may be listed in order of priority. Priorities may be set by weights, as discussed in more detail below. For example, the final product concentration may have a higher weight than the quality.
- each of the prioritization targets includes an extremum; three of the prioritization targets include a maximum and one prioritization target includes a minimum.
- Each of the prioritization targets also includes a target value (particularly the minimum product concentration value, the minimum specific productivity value, the maximum distance value and/or the minimum viability value, e.g., 3.5 g/L, 4.5 grams per cell per day, 65% or 15 units, as specified above, respectively).
- CS1_7 a clone referred to as CS1_7 is selected as the target clone.
- CS1_7 has a product concentration of 4.5 g/L, a specific productivity of 4.51 grams per cell per day, a quality of 14 and a final viability of 60%.
- Clone CS1_7 would be selected as the target clone according to conventional approaches, particularly because conventional approaches typically rely on at least one hard limit that leads to the exclusion of clones that do not meet the hard limit.
- hard limits can be automated using a decision tree. Further, multiple hard limits can be set by the user. However, use of one or more hard limits may lead to strict exclusion of clones that do not meet those limits and might not result in the selection of the best (i.e., optimal) target clone.
- a multivariate selection function (discussed in more detail below) used in the ranking accounts for each prioritization target and/or prioritization range, as well as weights for prioritization. Accordingly, a consistently ranked list of available clones is provided. Subjectivity is excluded and each ranking performed on the same data (even by different users) leads to a reliable and consistent result.
- a multivariate selection function is used to calculate scores for each one of the sets of clones from the correlated data according to the multivariate evaluation criteria.
- an objective function may be used to determine a distance between a process variable value (process parameter value and/or process output) and the prioritization target. More particularly, multiple process parameters and/or process outputs may be combined into a component, and the distance may be between the component and the prioritization target.
- the component may be derived using principle component analysis (PCA), however, other means suitable for calculating orthogonal components (i.e., vectors) from variables could also be used. Accordingly, components for each of the prioritization targets may be calculated.
- the components may be orthogonal (i.e., not correlated) and suitable for Euclidean distance calculations. Distance calculations may also be performed using partial least squares or orthogonal partial least square projections. According to one example, the distance between a candidate clone and the prioritization target may be calculated as follows:
- the objective function may include principal components (i.e., projection vectors) derived from process parameters and/or process outputs.
- the objective function may include a Euclidean distance calculation involving the principal (orthogonal) components.
- the combination of Euclidian distance and orthogonal components may be particularly advantageous, since possible correlations between variables (e.g., correlated glycans, as shown in Fig. 4 ) are reflected in the orthogonal components and an assumption that the variables are not correlated is unnecessary.
- the objective function may consider all process parameters and process outputs (i.e., if all are selected) or a subset (e.g., proper subset) of the process parameters and process outputs, as discussed in the example. Further, the absence of hard limits may prevent exclusion of optimal (e.g., the most efficient) clones.
- Step S105 may include determining the distance between 4.8 g/L and the target product concentration of 3.5 g/L. Determining the distance may involve normalizing the distance. In this case, since the product concentration is to be maximized, the inverse of the difference between the product concentration of clone CS2_2 and the target product concentration may be used.
- the distance is magnified in step S107.
- the selection function includes a magnifying function.
- the magnifying function may be a continuous non-linear function.
- an acceptable value may be within a prioritization range (e.g., an acceptable value of 4.5 within a range of 2-6) or between a target value and its corresponding extremum (e.g., an acceptable value of 4.5 having a target value of 3 and an extremum of maximize).
- a linear magnifying function (or no magnifying function) may result in clones that meet only a few of the targets (e.g., having relatively few acceptable values in between a target value and its corresponding extremum) being selected, if the few acceptable values (e.g., process outputs) are sufficiently close to the extremum in comparison to other clones.
- use of a linear function (or no function) could cause selection of clones that do not have acceptable values with respect to a relatively large number of targets. This may be undesirable.
- the magnifying function may also be referred to as a penalty function, since the magnifying function serves to increase the impact of the distance (i.e., impose a penalty according to the distance) between a value (e.g., process output) corresponding to a candidate clone and the prioritization target.
- a penalty function since the magnifying function serves to increase the impact of the distance (i.e., impose a penalty according to the distance) between a value (e.g., process output) corresponding to a candidate clone and the prioritization target.
- the magnifying function may be the same for all prioritization targets. In this way, the magnifying function can be used to influence values corresponding to a candidate clone (without consideration of others), while the weights can be used to prioritize values for corresponding to different prioritization targets against each other (e.g., by setting a weight for one prioritization target higher than a weight for another prioritization target).
- the magnified distance may be modified based on priority.
- the weights for prioritization may be used to modify the magnified distance.
- Each weight may be a value between 0 and 1 and the magnified distances may be modified by multiplying them by their corresponding weights.
- the modified distances may be aggregated to produce the score corresponding to the clone.
- the distances for all variables i.e., process parameters and process outputs
- the distances may be added together.
- clone CS2_2 being ranked higher (i.e., having a lower total distance from the prioritization target) than clone CS1_7.
- clone CS2_2 has a product concentration of 4.8 g/L, a specific productivity of 4.45 grams per cell per day, a quality of 8.0 and a final viability of 70%. Even though the specific productivity target of 4.5 grams per cell per day is not reached, the described approach results in the selection of clone CS2_2 as the target clone.
- clone CS2_2 has better values in targets other than specific productivity in comparison to the other candidate clones, clone CS2_2 would still have been discarded according to conventional approaches.
- Fig. 2 shows the selection of a set of target cells from multiple sets of candidate cells.
- the target cells and the candidate cells are clones, however, the described approach is applicable to other types of cells.
- a criteria filter pane 201 multivariate evaluation criteria for the selected process outputs are shown.
- the displayed prioritization targets include final viability (shown as a percentage with a target value of 65), final product concentration (shown in g/L with a target value of 3.5), final specific productivity (shown as "Qp", with a target value of 4.5), and quality (shown as "Distance”, with a target value of 15 units).
- the final viability, product concentration, and final specific productivity are to be maximized.
- the quality i.e., distance
- Weights for prioritization are also shown, with a weight of 0.4 for final viability, a weight of 1 for final product concentration, a weight of 0.8 for final specific productivity, and a weight of 0.8 for quality.
- a ranking pane 203 is also shown.
- the ranking pane 203 includes a first column for the candidate clones and a second column for the corresponding scores of the candidate clones.
- the clone CS2_2 (displayed as CS2-2) has the lowest score of 0.473 and therefore the highest rank.
- a clone/variable plot 205 shows each of the criteria for clone CS2_2 in relation to the other candidate clones.
- clone CS2_2 had the highest product concentration
- the final viability and final specific productivity values of CS2_2 were about average
- CS2_2 had a relatively low quality (i.e., distance).
- CS2_2 was given the highest rank.
- a raw data pane 207 shows a process trajectory for CS2_2 in comparison to process trajectories for the other clones.
- Fig. 3 shows how process parameter values and process outputs can be combined for evaluation.
- multivariate statistical techniques e.g., principal component analysis
- cell similarity indices may be calculated in both the quality domain e.g., using glycan profiles or spectral fingerprints, as well as in the process domain, where time series data can be combined and evaluated.
- Process outputs may correspond to the quality domain
- process parameters may correspond to the process data domain.
- partial least squares and/or orthogonal partial least squares may also be implemented for selecting target cells. Accordingly, an overview map may be presented, as shown in Fig. 3 , such that similarities between cells can be visualized and interpreted.
- a CS1 group of clones and a CS2 group of clones are displayed.
- the groups CS1 and CS2 may be delineated according to separate culture stations of the process control device.
- clones are referred to, the disclosed techniques are applicable to other types of cells.
- each point is the multivariate representation of the process parameters and the process outputs for one clone.
- the combination of process parameters and process outputs, as shown in Fig. 3 may highlight potential experimental problems not easily detected in the underlying data. Further, the depicted multivariate analysis may be useful in selecting the target clone.
- Multivariate statistics may be particularly useful when clones are represented using a set of process outputs (e.g., a glycan profile). Accordingly, the multivariate statistical analysis may provide similarity indices (e.g., principal components derived from process outputs), as shown in Fig. 3 , to be used in the ranking of sets of candidate clones. The similarity indices may be evaluated using the objective function, e.g., as discussed in the context of step S105 above.
- similarity indices e.g., principal components derived from process outputs
- the depicted example may also provide information on the correlation among different process parameters. This information can be applied when evaluating process parameters and process outputs against the prioritization targets, thereby improving the ranking. Further, correlation information may be useful for the user when setting values for prioritization targets or prioritization ranges.
- Fig. 3 data is collected from 24 processes. Each of the processes produces a distinct clone.
- the CS1 clones were produced in a first culture station and the CS2 clones were produced in a second culture station.
- Data collected from the processes carried out to produce the clones may be correlated and further evaluated as described in the context of Fig. 1 .
- Fig. 3 shows values with respect to two principal components, t[1] and t[2].
- further principal components may also be evaluated in addition to prioritization ranges, prioritization targets, and weights for prioritization as discussed in the context of Fig. 1 .
- the orthogonal components used in Fig. 3 may be more efficient for use in the ranking process in comparison to the many process parameters and process outputs. This is particularly the case because of conflicting (e.g., inversely correlated) targets, as discussed in more detail in the context of Fig. 4 .
- Fig. 4 shows process outputs (i.e., glycans) collected from a plurality of processes. More particularly, Fig. 4 shows a glycan profile derived from data collected from processes that produced the components of the 24 clones shown in Fig. 3 .
- targets i.e., prioritization targets
- the prioritization targets include an extremum, i.e., a minimum. Accordingly, it is desirable for both G1'f and G0f to have values as low as possible. However, since G1'f and G0f are inversely correlated, it is not possible to minimize both G1'f and G0f. Accordingly, one of these prioritization targets must be prioritized over the other.
- a display may be provided indicating an inverse correlation between at least two of these selected process outputs. Accordingly, an external input (such as by the user) may provide further prioritization targets before the scores are calculated via the multivariate selection function.
- the glycan profile for each clone is based on 12 process outputs.
- the glycan profile provides information regarding dependencies between different process outputs.
- each glycan may be considered a distinct process output and the glycan profile may show a correlation pattern among the glycans.
- Fig. 5 shows a projection plot depicting a principal component transformation of the 12 glycan profile variables depicted in Fig. 4 . Accordingly, a value for each clone is displayed and the value is derived from the 12 glycan profile variable values for that clone.
- the principal components depicted in Fig. 5 are an example of components (i.e., projection vectors t i ) that can be used in the context of the objective function discussed in connection with Fig. 1 . Further, the components depicted in Fig. 5 may be used to calculate the scores via the multivariate selection function.
- the x axis depicted in Fig. 5 represents 98% of the total variation in the 12 glycan profile variables.
- 98% of the variation in the 12 glycan profile process outputs is one-dimensional. While combining process outputs into principal components may make calculation of the scores via the multivariate selection function more efficient, it would also be possible to use process parameters and/or process outputs directly, without the additional step of determining principal components.
- the origin of the projection plot is a score corresponding to a reference clone.
- the reference clone may be represented by a plurality of target values.
- Fig. 6 shows functionality of a tool that may be used to implement the method for selecting at least one set of target cells from multiple sets of candidate cells, as discussed above.
- the tool may be implemented in hardware and/or software.
- the tool may be implemented in the process control device mentioned above and depicted in Figs. 7 and 8 .
- steps S601 to S609 there may be some overlap between steps S601 to S609.
- actions carried out in step S601 may be carried out after actions carried out in step S603.
- some of the results of step S601 may be used in steps subsequent to step S603, without the processing carried out in step S603. Corresponding considerations may apply to the other steps.
- Step S601 data is imported.
- Step S601 may include receiving data collected from a plurality of processes, wherein each process produces a distinct set of candidate cells.
- the imported data may include process data.
- Process data may be considered synonymous with process parameter values.
- Process data may include time dependent data sampled from the processes. Examples of process data are pH, product titer, viable cell density, glucose, dissolved oxygen, and/or oxygen consumption.
- quality data may be imported.
- Quality data may be understood as process outputs.
- the quality data may describe the end quality of the candidate cells. More particularly, the quality data may describe the cell line, cell strain, or clone processed.
- Typical quality data may be glycosylation patterns, charge variants, aggregates, low molecular weight species, and/or glycan residues displayed as a profile (e.g., as profile vectors).
- Process outputs may also include aggregated process data. For example, viable cell density may be measured throughout the process and the measurements may be received as process parameter values. A final viable cell density may be received as a process output of a process.
- Step S601 may also include handling missing data. This step may be carried out in the context of correlating the received data. Missing data may include data that is not collected or sampled frequently enough. For example, glucose and/or lactate may be sampled only once per day, however, a more complete picture of glucose and/or lactate levels may be desired. Accordingly, it may be desirable to simulate hourly measurement of glucose or lactate by filing in data for missing samples.
- the missing data may be filled in using mechanistic modeling procedures and/or multivariate prediction (e.g., partial least squares regression).
- Mechanistic modeling and multivariate prediction models may also be used to predict future behavior of candidate cells. Accordingly, mechanistic modeling may provide input on the biological state of candidate cells at any given time and may enable early evaluation of candidate cells.
- Prediction of future behavior may make it possible to determine what would happen if processing of candidate cells is terminated prematurely, e.g., due to an infection. For example, if processing of candidate cells is terminated prematurely, prediction of future behavior of the candidate cells may still enable process data for those candidate cells to be incorporated into calculations performed via the multivariate selection function in order to arrive at a score for the candidate cells that can be used in the ranking discussed above.
- Mechanistic modeling can also be used to improve the quality of measured data by verifying and correcting values that violate known metabolic ratios, thereby ensuring that higher quality data is used in the ranking.
- Mechanistic modeling may also be used to estimate process parameter values (e.g., cell death rate) that are difficult or impossible to measure directly and thereby add further viable information that can be used when calculating scores for each of the candidate cells.
- viable cell density measurements may have an error of +/- 10%. Smoothing of the measurements may be carried out in order to exclude the erroneous measurements, e.g., process parameter values or process outputs.
- Step S601 may also include visual quality control. Accordingly, the received data may be graphically presented such that outliers can be easily identified and excluded. Further, data can be corrected as desired.
- Step S601 may also involve data exclusion. More particularly, at least one of the processes may fail. In particular, events such as contamination and/or system failure, unexplained or inconsistent biological factors, or human error may lead to failure of the process. Data from the failed process may be excluded.
- Step S601 may also include grouping the received data. For example, replicates, minipools, or biosimilar cells (e.g., clones) may be grouped for analysis. Prioritization ranges and/or prioritization targets may be set for each group, or for the entire set of cells.
- Step S601 may also include data matching.
- data may be received from various sources.
- the data may be matched and synchronized for analysis.
- the sources may include the process control device itself, and possibly an external analysis device.
- External analysis devices may include one or more of the following: a device for offline spectroscopy measurement (spectrometer), a device for inline (i.e., online, in situ) biomass measurement, a device for nutrient and metabolite measurement.
- the spectrometer may be an apparatus to separate (subatomic) particles, atoms, and molecules by their mass, momentum, or energy). For example, spectroscopic measurements may be carried out using an Acquity iClass UPLC and Xevo TQS triple quadrupole mass spectrometer (Waters, Milford, MA). Other devices may also be used.
- the device for nutrient and metabolite measurement may measure parameters online.
- nutrients include glucose and lactate.
- metabolites include methanol and ethanol.
- the device may perform up to 60 analyses per hour for a filtration probe and up to 30 analyses per hour via a dialysis setup. More specifically, glucose and lactate may be analyzed using a Bioprofile flex (Nova Biomedical Corporation, Waltham, MA).
- Viable cell concentration may be analyzed using a cell viability analyzer, such as the Vi-Cell Automated cell viability analyzer (Beckman Coulter, Brea, California). Other devices may also be used.
- a cell viability analyzer such as the Vi-Cell Automated cell viability analyzer (Beckman Coulter, Brea, California). Other devices may also be used.
- Step S601 may also include creating a project.
- the project may provide a framework to use and store multivariate evaluation criteria, as discussed in more detail below.
- step S603 criteria for the selected process parameters and/or the selected process outputs may be set.
- step S603 may include selecting a (proper) subset of the process parameters and/or a (proper) subset of the process outputs. Accordingly, the selection of the process parameters may exclude one or more of the process parameters. The selection of the process outputs may exclude one or more of the process outputs. The selected process parameters and/or process outputs may then be received, e.g., stored by, the process control device.
- Step S603 may include displaying the received data collected from the plurality of processes. Further, the process parameters and process outputs may be displayed.
- Data may be displayed in the form of a table.
- process parameter values may be visualized in a data table.
- the table may facilitate correction of obvious errors in the data.
- Display of the data in the data table may facilitate exclusion of candidate cells or identification of one or more sets of candidate cells as reference cells. For example, if one of the sets of candidate cells exhibits an abnormal profile, the set of candidate cells may be eliminated.
- Acceptability ranges for process outputs may be set.
- the acceptability ranges may be set so as to exclude one or more outliers.
- Setting acceptability ranges i.e., acceptable ranges
- Setting acceptability ranges may be part of a prefiltering process. To aid in the setting of acceptability ranges, an overview display of the received data may be provided. Accordingly, it can be easily determined how much of the received data is excluded by setting an acceptability range. If the acceptability range is set too strictly, the number of sets of candidate cells that pass through the prefiltering process may be too limited.
- an acceptability range is set to filter out sets of candidate cells having a low titer
- the absence of high titer producing sets of candidate cells among all the sets of candidate cells may limit the number of sets of candidate cells that are passed through the prefiltering process. This may be undesirable.
- Step S603 may also include a raw data visualization.
- the raw data visualization may help facilitate understanding of which data is excluded by the specified acceptability ranges.
- Step S605 may include receiving multivariate evaluation criteria for the selected process parameters and/or the selected process outputs.
- the multivariate evaluation criteria may include aggregating or further processing at least one of the selected process parameters and/or the selected process outputs.
- the multivariate evaluation criteria may include weights for prioritization.
- the multivariate evaluation criteria may include prioritization ranges and/or targets, wherein each target is an extremum and/or a target value.
- Prioritization targets may be set for a (proper) subset of the selected process parameters and/or process outputs.
- the extremum may be to maximize or minimize.
- the prioritization target may include a target value or set point.
- the target value may be a specific reference or limit value.
- Prioritization ranges may be received as input from the user. However, the prioritization ranges may be modified when calculating the scores for each one of the sets of candidate cells. Weights for prioritization may be used to define the importance of each of the prioritization targets. For example, a weight of 1.0 may be given to the most important prioritization target. Weights close to 0 may be given to relatively unimportant prioritization targets or prioritization ranges.
- Step S605 may include calculating, via the multivariate selection function, scores for each one of the sets of candidate cells from the correlated data according to the multivariate evaluation criteria.
- the multivariate selection function may include an objective function, more particularly a cost function.
- the objective function may be referred to as a desirability function.
- the objective function may be used to rank the candidate cells according to how well they fit the multivariate evaluation criteria.
- the objective function may be non-linear.
- the objective function may be exponential, e.g., quadratic.
- the objective function may quantify the distance from a numerical target (e.g., a prioritization target or range) and aggregate penalties based on the evaluation criteria and the weights for prioritization.
- the numerical target may be a set of reference cells having a biosimilar definition in comparison to the sets of candidate cells.
- the objective function may be as specified in the discussion of step S105 of Fig. 1 above.
- Calculation of scores for each one of the sets of candidate cells may be performed multiple times (i.e., iteratively) using different evaluation criteria, particularly because it may be determined that two or more of the evaluation criteria are inversely correlated, as shown in Fig. 4 .
- the multivariate selection function may produce a score that reflects how well a set of candidate cells fulfills the multivariate evaluation criteria.
- the score may be referred to as a desirability index.
- Selected profiles for candidate cells can be visualized graphically and enable the inspection of cell profiles with scores that are close to each other, as shown in Fig. 2 . For example, if a set of candidate cells is ranked fifth, e.g., in the ranking 203, the corresponding variable plots for the set of candidate cells and other sets of candidate cells may facilitate determination of how the ranking was generated. Such comparisons may enable reprioritization and guide further selection iterations. Accordingly, the display of graphical data, e.g., as shown in Fig. 2 , may facilitate analysis that would not be possible from the underlying raw data. Moreover, the number of selection iterations may be reduced and require less user intervention in comparison to conventional approaches.
- a multivariate correlation analysis tool e.g., the tool implementing the method for selecting at least one set of target cells for multiple sets of candidate cells, can help the user to adjust the selection process, including prioritization targets, weights, and ranges, in order to ensure that the optimal set of target cells is selected. Further, the selection of sets of candidate cells discussed above may reflect both process parameter values and process outputs.
- the evaluation criteria can be saved for later use with other sets of candidate cells or shared with other users. Use of the same multivariate evaluation criteria by multiple users for different evaluations may ensure consistency. In particular, the same criteria may be applied to different data sets by different users. This may eliminate the subjectivity that is often present in conventional approaches. Further, the same criteria may be used for different batches of data from the same project.
- step S607 analysis may be carried out. As noted above, some of the actions carried out in the context of step S607 may be carried out in combination with or even before steps carried out in the context of step S605.
- At least one of the prioritization targets may be calculated. More particularly, at least one of the target values may be calculated. For example, integrals and/or averages of the process parameter values, relations, and predictions based on process history may be calculated. Further, some process outputs may also be calculated. Calculated and/or predicted values may provide more stable input for the ranking of the sets of candidate cells and the selection of at least one of the sets of candidate cells as the target cells.
- mechanistic modeling algorithms may be used to fill in missing data and predict process trajectories and/or final results, as discussed above. For example, candidate cells that stopped growing earlier than expected may be compared with other candidate cells that continued producing results until the end of their corresponding processes. Process data and quality data for the candidate cells that stopped growing earlier than expected may be extrapolated or interpolated from the other candidate cells that continued producing results until the end of their corresponding processes. The trajectory predictions for candidate cells that terminated early may provide a more complete set of data that can be used to improve the ranking of the sets of candidate cells.
- multivariate modeling may be used to compensate for errors in measurement due to sampling, handling or inherent flaws in external analytic devices.
- the multivariate modeling may use fundamental bioprocessing correlations to produce more reliable data. This may be part of the correlation carried out after data is received or may be carried out in a further iteration of the described method after an initial ranking of the sets of candidate cells has been produced.
- the score calculated via the multivariate selection function may be based on all selected process parameters and all selected process outputs. Further, information from the process trajectory of each process may also be used. Scores calculated via the multivariate selection function may be compared to a prioritization target (e.g., biosimilar cells or a group value).
- Comparisons can be made within a group or to a specific reference process (i.e., a process that produces reference cells). These comparisons may facilitate setting different prioritization ranges, targets and weights in further iterations of the above described method. Further, process trajectories for each of the sets of candidate cells can be used to calculate distances between sets of candidate cells, groups, or from target values for use in the ranking of the sets of candidate cells. In addition to process trajectories, other multivariate criteria may be used in the ranking of candidate cells, particularly other quality data such as glycan profiles.
- Step S607 may also include correlation analysis, as discussed in connection with Figs. 4 and 5 .
- the correlation analysis may give information on how to improve candidate cell selection, since some of the received evaluation criteria may contradict other criteria (e.g., the criteria may include inversely correlated targets). Understanding the correlation between different variables (i.e., process parameters and process outputs) may facilitate tuning of the tool in order to arrive at an optimal selection of target cells, e.g., in a minimal number of iterations.
- Selected candidate cells may be compared with all other candidate cells for analysis.
- Multivariate approaches such as principal component analysis (PCA) and partial least squares (PLS) regression can be used for visual comparison and calculation of distance between different sets of candidate cells.
- Graphic representation and visualization of the results may help the user determine similarities between specific sets of candidate cells or groups.
- Components describing individual sets of candidate cells may be displayed as dots in two-dimensional diagrams, as shown in Figs. 3 and 5 , with a possibility to color or set sizes according to any variable or preference used in the selection process.
- the coloring and sizing may be part of a three-dimensional graph.
- a report may optionally be produced.
- the report may include a document reporting one or more selection results, the multivariate evaluation criteria, prioritization targets, target values and/or extrema, rankings, statistical correlations, and/or observations.
- the report may provide information supporting (i.e., reasons for) the selection of the set of target cells from the multiple sets of candidate cells. If multiple iterations of the method have been carried out, the report may include a summary of results from each iteration.
- Fig. 7 shows an example of how mechanistic modeling can be used to smooth process outputs.
- the example of Fig. 7 depicts "E5" cells, i.e., a particular type of candidate cell.
- VCD measurements in cells/mL are denoted by “x” marks and cell viability measurements (indicating the percentage of cells that are viable, i.e., still alive) are denoted by filled in circles.
- the VCD measurements include an initial VCD measurement of 2.38827.
- the cell viability measurements include an initial cell viability measurement close to 100%. VCD in cells/mL and cell viability in percentage are shown on the vertical axis and time in days is shown on the horizontal axis.
- a VCD curve 701 and a cell viability curve 703 are calculated from the measurements and cell-specific constants "K" corresponding to the cells.
- the constants "K” are specific to the candidate cells depicted and differ for other (different) sets of candidate cells.
- X total total number of cells, alive and dead
- X VCD viable cell density
- X dead number of dead cells
- u max , K dead , K toxic , and K inhibit are cell-specific constants; equations (1) to (3) may be solved to derive the cell-specific constants according to a conventional optimization method.
- u max maximum growth rate of the cells
- K toxic increase in death rate due to environmental toxicity brought about by dead cells
- K dead cell death rate
- K inhibit coefficient characterizing a reduction in growth rate due to the total number of cells
- K inhibit reflects the principle that cells grow more slowly when the total number of cells is greater. Thus, K inhibit may grow with cell density (i.e., cells may be inhibited from growing as cell density increases). Equations (1) to (3) reflect the effects of cell density on cell growth, but other effects may also be considered.
- Mechanistic modeling may be carried out during correlation of received data in order to fill in missing measurements, e.g., by a plotting one or more curves (as shown in Figure 7 ) based on existing measurements and known physical characteristics of the cells. Accordingly, the mechanistic modeling may affect values of the selected process parameters and selected process outputs.
- cell-specific constants in the context of correlation (e.g., filling in missing data, data smoothing or interpolation) of process parameter values and process outputs via mechanistic modeling (as described above) has the effect of reflecting physical properties of cells when carrying out the correlation. This may lead to more accurate calculation of the scores for the sets of candidate cells (particularly in comparison to approaches that rely exclusively on multivariate statistical approaches such as PCA or PLS), and accordingly, selection of the optimal target cells from the sets of candidate cells.
- a process control device 10 (also referred to as a bioreactor system) including an array of vessels 100 (e.g., microscale bioreactors) is shown in Fig. 8 .
- the process control device 10 may be mounted to the deck of a base station in a larger scale process control device.
- the process control device 10 may be a microscale process control device suitable for mounting to a macroscale process control device.
- the macroscale process control device may include vessels having a size that differs from a size of the vessels of the microscale process control device by at least one order of magnitude.
- the process control device 10 comprises a base 12, to which is mounted a base plate 13 defining a receiving station 14 for removably receiving a plurality of vessels 100.
- a clamp plate (not shown) may be removably connected to the base plate 13, in a position overlaying the receiving station 14, via a pair of posts 22 projecting from the upper surface of the base plate 13.
- the clamp plate may facilitate a drive connection between the drive mechanism of a stirrer 116 (described below).
- the receiving station 14 can hold up to twelve vessels 100 in two rows of six at respective locations 16.
- Fig. 8 six vessels 100 are shown in position in their respective vessel receiving locations 16, while six of the vessel receiving locations 16 are shown empty to better illustrate fluid ports 314 a-c in the base plate 13.
- the receiving station 14 could be designed to accommodate a greater or lesser number of vessels 100 and the vessels 100 could be arranged in any suitable configuration.
- One or more heaters or chillers may be located adjacent to the vessel receiving locations 16 to control the temperature of the vessels 100.
- one of the vessels 100 comprises a chamber 105 for receiving a fluid 107 (e.g., a cell culture solution) having a headspace 109 above.
- the vessel 100 includes a pipette access port 106, to which a cap 108 is removably attached. The cap 108 is removed for fluids to be pipetted into or out of the vessel 100.
- a fluid input port 112 may include a filter 114.
- the stirrer 116 comprising blades 118, may be rotatably mounted at the base of a vertical shaft 120 within the vessel 100.
- the upper end of the vertical shaft 120 includes a drive input 124 (e.g., for the drive mechanism, not shown).
- a pH sensor spot 126 and a dissolved oxygen (DO) sensor spot 128 are disposed at the bottom of the vessel 100, such that they are able to detect the pH and DO levels of the fluid 107 and to be interrogated from the exterior of the vessel 100.
- DO dissolved oxygen
- Venting of the vessel chamber 105 is achieved via a labyrinthine path connecting the chamber 105 to the atmosphere via the stirrer shaft drive input 124.
- a separate vent port may be provided towards the top of the vessel 100.
- a lip 130 may project out to the side of the vessel 100.
- the lip 130 includes a through port 132b (two optional additional ports 132a and 132c are not shown).
- a gallery plate 134 is secured above a portion of the top of the vessel 100.
- the gallery plate 134 includes at least one groove 136b extending to the fluid input port 112 at the top of at least one tube 110b.
- the gallery plate further includes at least one through port 132b.
- the lip 130 and the gallery plate 134 together form a rigid ledge projecting to the side of the vessel.
- the clamp plate (not shown) may also reinforce a seal between the through port 132b and the fluid ports 314a-c.
- a valve assembly 300 is mounted to the underside of the base 12. The valve assembly is received in a cavity of the base station when the process control device 10 is connected to the base station.
- the process control device 10 is loaded with vessels 100, each vessel being placed in a respective vessel receiving location within the receiving station 14.
- the port 132b in the bottom surface of the lip 130 is aligned with and forms a sealed connection with the corresponding receiving station fluid port 314b on the upper surface of the base plate 13.
- the respective ports are automatically aligned with one another on insertion by virtue of the defined locations of the vessel receiving station, including the fluid ports 314a-c adjacent thereto, and the rigid ledge, which places the corresponding vessel connection ports 132a-c in alignment with the receiving station fluid ports 314a-c.
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- Medical Informatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Public Health (AREA)
- Bioethics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biotechnology (AREA)
- Epidemiology (AREA)
- General Health & Medical Sciences (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Apparatus Associated With Microorganisms And Enzymes (AREA)
- Measuring Or Testing Involving Enzymes Or Micro-Organisms (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Priority Applications (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP18200386.3A EP3640946A1 (de) | 2018-10-15 | 2018-10-15 | Multivariatischer ansatz für die auswahl der biologischen zelle |
US17/285,459 US20210383893A1 (en) | 2018-10-15 | 2019-10-07 | Multivariate approach for cell selection |
PCT/EP2019/077040 WO2020078747A1 (en) | 2018-10-15 | 2019-10-07 | Multivariate approach for cell selection |
CN201980061150.4A CN112714934B (zh) | 2018-10-15 | 2019-10-07 | 细胞选择的多变量方法 |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP18200386.3A EP3640946A1 (de) | 2018-10-15 | 2018-10-15 | Multivariatischer ansatz für die auswahl der biologischen zelle |
Publications (1)
Publication Number | Publication Date |
---|---|
EP3640946A1 true EP3640946A1 (de) | 2020-04-22 |
Family
ID=64082855
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
EP18200386.3A Pending EP3640946A1 (de) | 2018-10-15 | 2018-10-15 | Multivariatischer ansatz für die auswahl der biologischen zelle |
Country Status (4)
Country | Link |
---|---|
US (1) | US20210383893A1 (de) |
EP (1) | EP3640946A1 (de) |
CN (1) | CN112714934B (de) |
WO (1) | WO2020078747A1 (de) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2021240103A1 (fr) * | 2020-05-29 | 2021-12-02 | Orange | Regroupement de trajectoires dans le domaine chiffre |
WO2022066418A1 (en) * | 2020-09-24 | 2022-03-31 | Amgen Inc. | Methods and systems for determining a minimum number of cell line clones necessary to produce a product having a set of target product attributes |
CN117793764A (zh) * | 2023-12-27 | 2024-03-29 | 广东宜通衡睿科技有限公司 | 5g专网软探针拨测数据完整性校验和补全方法及系统 |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20230137741A1 (en) | 2021-11-01 | 2023-05-04 | Matterworks Inc | Methods and compositions for analyte quantification |
US20240112750A1 (en) * | 2022-10-04 | 2024-04-04 | BioCurie Inc. | Data-driven process development and manufacturing of biopharmaceuticals |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6673532B2 (en) * | 2000-08-14 | 2004-01-06 | University Of Maryland, Baltimore County | Bioreactor and bioprocessing technique |
WO2007018868A1 (en) * | 2005-07-25 | 2007-02-15 | Biogen Idec Ma Inc. | System and method for bioprocess control |
US20080133044A1 (en) * | 2004-08-23 | 2008-06-05 | Herbert Grieb | Method and Arrangement For the Online Regulation of a Batch Process in a Bioreactor |
US20100191361A1 (en) * | 2009-01-23 | 2010-07-29 | Mks Instruments, Inc. | Controlling a Manufacturing Process with a Multivariate Model |
EP2270129A2 (de) * | 2009-07-01 | 2011-01-05 | The Automation Partnership (Cambridge) Limited | Bioreaktorsystem und zugehörige Verfahren zur Verarbeitung von Bioreaktorgefäßen |
EP2447717A1 (de) * | 2010-10-27 | 2012-05-02 | Lonza Biologics plc. | Schnelles Verfahren zur gezielten Zellenselektion (Zelllinienselektion) |
US20130268238A1 (en) * | 2012-04-06 | 2013-10-10 | Mks Instruments, Inc. | Multivariate Monitoring of a Batch Manufacturing Process |
US20140136146A1 (en) * | 2012-04-06 | 2014-05-15 | Mks Instruments, Inc. | Multivariate prediction of a batch manufacturing process |
WO2017174580A1 (en) * | 2016-04-04 | 2017-10-12 | Boehringer Ingelheim Rcv Gmbh & Co Kg | Real time monitoring of product purification |
US20170355947A9 (en) * | 2014-07-02 | 2017-12-14 | Biogen Ma Inc. | Cross-scale modeling of bioreactor cultures using raman spectroscopy |
Family Cites Families (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1520557A (zh) * | 1999-09-03 | 2004-08-11 | 产品生产过程中的参数值优化方法 | |
US6876931B2 (en) * | 2001-08-03 | 2005-04-05 | Sensys Medical Inc. | Automatic process for sample selection during multivariate calibration |
CN1647067A (zh) * | 2002-04-04 | 2005-07-27 | 石原产业株式会社 | 数据解析装置及方法 |
WO2005017646A2 (en) * | 2002-08-06 | 2005-02-24 | The Johns Hopkins University | System, software and methods for biomarker identification |
US8065089B1 (en) * | 2004-03-30 | 2011-11-22 | University Of North Carolina At Charlotte | Methods and systems for analysis of dynamic biological pathways |
US20070100475A1 (en) * | 2005-10-25 | 2007-05-03 | Korchinski William J | Method and apparatus for applying reduced nonlinear models to optimization of an operation |
CN102509243A (zh) * | 2011-09-20 | 2012-06-20 | 河北中烟工业有限责任公司 | 卷烟制造过程质量评价方法及其系统 |
PL2936124T3 (pl) * | 2012-12-20 | 2017-08-31 | F.Hoffmann-La Roche Ag | Sposoby oceniania medycznych krzywych pomiarowych |
CN105408746A (zh) * | 2013-02-28 | 2016-03-16 | 普罗吉涅股份有限公司 | 基于图像的人胚胎细胞分类的装置、方法和系统 |
US10762167B2 (en) * | 2013-09-27 | 2020-09-01 | Varian Medical Systems International Ag | Decision support tool for choosing treatment plans |
CN106471508A (zh) * | 2014-06-20 | 2017-03-01 | 康涅狄格儿童医疗中心 | 自动化细胞培养系统及对应的方法 |
CN107111874B (zh) * | 2014-12-30 | 2022-04-08 | 文塔纳医疗系统公司 | 用于共表达分析的系统和方法 |
JP6791598B2 (ja) * | 2015-01-22 | 2020-11-25 | ザ ボード オブ トラスティーズ オブ ザ レランド スタンフォード ジュニア ユニバーシティー | 異なる細胞サブセットの比率の決定方法およびシステム |
CN104598774A (zh) * | 2015-02-04 | 2015-05-06 | 河南师范大学 | 基于logistic与相关信息熵的特征基因选择方法 |
DE102015110893B3 (de) * | 2015-07-06 | 2016-03-24 | Sartorius Stedim Biotech Gmbh | Verfahren zum Messen einer Mehrzahl von Zustandsparametern eines in einem Behälter enthaltenen Fluids |
US10013654B1 (en) * | 2017-11-29 | 2018-07-03 | OJO Labs, Inc. | Cooperatively operating a network of supervised learning processors to concurrently distribute supervised learning processor training and provide predictive responses to input data |
CN108490782B (zh) * | 2018-04-08 | 2019-04-09 | 中南大学 | 一种基于选择性双层集成学习适用于复杂工业过程产品质量指标缺失数据补全的方法及系统 |
-
2018
- 2018-10-15 EP EP18200386.3A patent/EP3640946A1/de active Pending
-
2019
- 2019-10-07 WO PCT/EP2019/077040 patent/WO2020078747A1/en active Application Filing
- 2019-10-07 CN CN201980061150.4A patent/CN112714934B/zh active Active
- 2019-10-07 US US17/285,459 patent/US20210383893A1/en active Pending
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6673532B2 (en) * | 2000-08-14 | 2004-01-06 | University Of Maryland, Baltimore County | Bioreactor and bioprocessing technique |
US20080133044A1 (en) * | 2004-08-23 | 2008-06-05 | Herbert Grieb | Method and Arrangement For the Online Regulation of a Batch Process in a Bioreactor |
WO2007018868A1 (en) * | 2005-07-25 | 2007-02-15 | Biogen Idec Ma Inc. | System and method for bioprocess control |
US20100191361A1 (en) * | 2009-01-23 | 2010-07-29 | Mks Instruments, Inc. | Controlling a Manufacturing Process with a Multivariate Model |
EP2270129A2 (de) * | 2009-07-01 | 2011-01-05 | The Automation Partnership (Cambridge) Limited | Bioreaktorsystem und zugehörige Verfahren zur Verarbeitung von Bioreaktorgefäßen |
EP2447717A1 (de) * | 2010-10-27 | 2012-05-02 | Lonza Biologics plc. | Schnelles Verfahren zur gezielten Zellenselektion (Zelllinienselektion) |
US20130268238A1 (en) * | 2012-04-06 | 2013-10-10 | Mks Instruments, Inc. | Multivariate Monitoring of a Batch Manufacturing Process |
US20140136146A1 (en) * | 2012-04-06 | 2014-05-15 | Mks Instruments, Inc. | Multivariate prediction of a batch manufacturing process |
US20170355947A9 (en) * | 2014-07-02 | 2017-12-14 | Biogen Ma Inc. | Cross-scale modeling of bioreactor cultures using raman spectroscopy |
WO2017174580A1 (en) * | 2016-04-04 | 2017-10-12 | Boehringer Ingelheim Rcv Gmbh & Co Kg | Real time monitoring of product purification |
Non-Patent Citations (1)
Title |
---|
LI B. ET AL: "Performance monitoring of a mammalian cell based bioprocess using Raman spectroscopy", ANALYTICA CHIMICA ACTA, ELSEVIER, AMSTERDAM, NL, vol. 796, 6 August 2013 (2013-08-06), pages 84 - 91, XP028711483, ISSN: 0003-2670, DOI: 10.1016/J.ACA.2013.07.058 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2021240103A1 (fr) * | 2020-05-29 | 2021-12-02 | Orange | Regroupement de trajectoires dans le domaine chiffre |
FR3111046A1 (fr) * | 2020-05-29 | 2021-12-03 | Orange | Regroupement de trajectoires dans le domaine chiffré |
WO2022066418A1 (en) * | 2020-09-24 | 2022-03-31 | Amgen Inc. | Methods and systems for determining a minimum number of cell line clones necessary to produce a product having a set of target product attributes |
CN117793764A (zh) * | 2023-12-27 | 2024-03-29 | 广东宜通衡睿科技有限公司 | 5g专网软探针拨测数据完整性校验和补全方法及系统 |
Also Published As
Publication number | Publication date |
---|---|
WO2020078747A1 (en) | 2020-04-23 |
US20210383893A1 (en) | 2021-12-09 |
CN112714934B (zh) | 2024-05-10 |
CN112714934A (zh) | 2021-04-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
EP3640946A1 (de) | Multivariatischer ansatz für die auswahl der biologischen zelle | |
US20220228102A1 (en) | Data-driven predictive modeling for cell line selection in biopharmaceutical production | |
US11542564B2 (en) | Computer-implemented method, computer program product and hybrid system for cell metabolism state observer | |
EP2594944B1 (de) | Automatische analysevorrichtung, analyseverfahren und informationsverarbeitungsvorrichtung | |
CN102246047B (zh) | 自动分析装置 | |
EP2428802B1 (de) | Automatische analysevorrichtung und analyseverfahren dafür | |
AU2013235400B2 (en) | Method and system for detection of microbial growth in a specimen container | |
US20200202051A1 (en) | Method for Predicting Outcome of an Modelling of a Process in a Bioreactor | |
CN111201434A (zh) | 用于控制细胞培养物中的过程变量的原位拉曼光谱系统和方法 | |
US20220146987A1 (en) | Multivariate process chart to control a process to produce a chemical, pharmaceutical, biopharmaceutical and/or biological product | |
CN114127549A (zh) | 分析装置以及分析方法 | |
US20220137579A1 (en) | Optimized process control | |
US20220291247A1 (en) | Walk-away time visualization methods and systems | |
JP7299914B2 (ja) | 異常判定方法、および自動分析装置 | |
US20200291341A1 (en) | Pseudo sample creation method and apparatus, and numerical model creation system and method | |
WO2021110520A1 (en) | Adapting control of a cell culture in a production scale vessel with regard to a starting medium | |
Ramm et al. | Systematic Preprocessing of Dielectric Spectroscopy Data and Estimating Viable Cell Densities | |
CA3195491A1 (en) | Methods and systems for determining a minimum number of cell line clones necessary to produce a product having a set of target product attributes | |
Herold | Automatic generation of process models for fed-batch fermentations based on the detection of biological phenomena | |
Pedersen | Predictive Modeling of Pipetting Dynamics | |
Munack | Measurement, modelling and control problems for biotechnical processes |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE APPLICATION HAS BEEN PUBLISHED |
|
AK | Designated contracting states |
Kind code of ref document: A1 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
|
AX | Request for extension of the european patent |
Extension state: BA ME |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE |
|
17P | Request for examination filed |
Effective date: 20201015 |
|
RBV | Designated contracting states (corrected) |
Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
|
RIN1 | Information on inventor provided before grant (corrected) |
Inventor name: YAU-ROSE, SIN-YEE Inventor name: VIKSTROEM, ERNST CONNY Inventor name: RAENNAR, NILS ERIK STEFAN Inventor name: ZEHE, CHRISTOPH Inventor name: JOHANSSON, ERIK AXEL Inventor name: MCCREADY, CHRISTOPHER PETER |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: EXAMINATION IS IN PROGRESS |
|
17Q | First examination report despatched |
Effective date: 20230125 |
|
P01 | Opt-out of the competence of the unified patent court (upc) registered |
Effective date: 20230508 |